Skip to content

Erdős Problem #322

Editorial depth: Reviewed explanatory article — evidence qualified.
Snapshot classification: open (not automatically the live result).
Subject areas: number theory, powers.

Read this status correctly

The status above is reproduced from GCL's immutable external-catalogue snapshot. It is not a new claim of current mathematical resolution, a source-semantic review, an independent proof check, or GCL certification. A published proof and a GCL replay are separate events.

The question in plain language

Erdős Problem 322 asks how often an integer can be written as a sum of exactly \(k\) many kth powers, when \(k\geq 3\). For a fixed \(k\), let

\[ r_k(n)=\#\{(x_1,\ldots,x_k)\in\mathbb N_{>0}^k:x_1^k+\cdots+x_k^k=n\}. \]

The tuples are ordered: changing the order of the same summands counts again. The problem asks for the order of growth of \(r_k(n)\) and, more specifically, whether there is a constant \(c>0\) such that \(r_k(n)>n^c\) for infinitely many \(n\). This is an asymptotic question for each fixed exponent, not a request to find one unusually representable integer.

The original issue identifies the problem as an open, source-audit assignment. Its pinned packet points to the maintained Problem 322 page, and to teorth/erdosproblems commit 536408716724a2402bdfa7c1d764bf83d64a8c89. In that commit, the relevant data/problems.yaml entry has Git blob 4e60263a80af6207aca494d2e1388e44fa372889, status open, and last update 2025-08-31.

A small example

Take \(k=3\) and \(n=10\). Any positive base \(x\) occurring in a representation satisfies \(x^3\leq 10\), so \(x\) is either \(1\) or \(2\). The exact possibilities are therefore

\[ 1^3+1^3+2^3=10, \]

with the larger summand in any one of the three positions. Thus

\[ r_3(10)=3, \]

and the ordered triples are \((1,1,2)\), \((1,2,1)\), and \((2,1,1)\). The following exact-arithmetic enumeration returns precisely those three triples:

from itertools import product
sol = [x for x in product(range(1, 11), repeat=3) if sum(t**3 for t in x) == 10]
print(sol, len(sol))

Its output is [(1, 1, 2), (1, 2, 1), (2, 1, 1)] 3; no floating-point approximation is involved. This verifies the definition at one finite input. It says nothing about the order of \(r_3(n)\) as \(n\) grows, nor about infinitely many \(n\) with a fixed positive-power lower bound.

The example also makes the convention important. The 1936 primary paper defines its function using nonnegative bases, so zero-containing triples are included there. The positive-base convention above is the one used in the current formal file. These conventions agree only after an explicit comparison, not by notation alone.

What is known

The primary source is Paul Erdős, “On the Representation of an Integer as the Sum of k k-th Powers,” Journal of the London Mathematical Society 11 (1936), 133–136. An accessible scan shows on its first page that \(f(m)\) counts representations by kth powers of nonnegative integers. It recalls Hardy and Littlewood’s Hypothesis K, namely an upper bound of the form \(f(m)=O_\epsilon(m^\epsilon)\) for every \(\epsilon>0\), and states the opposite-direction result

\[ f(m)>\exp\left(\frac{c_1\log m}{\log\log m}\right) \]

for infinitely many \(m\), with a positive constant depending on \(k\). The proof occupies the remainder of the paper, including separate odd- and even \(k\) arguments. In the normalization \(\exp(c_1\log m/\log\log m)=m^{c_1/\log\log m}\), this grows faster than every fixed power of \(\log m\) but slower than \(m^\delta\) for any fixed \(\delta>0\).

The same paper records that Mahler had just shown Hypothesis K to be false for \(k=3\), with many representations along a sequence of twelfth powers. The maintained page also attributes the corresponding lower-bound result independently to Erdős and Chowla, and records Mahler’s \(k=3\) result, the weaker Hardy–Littlewood Hypothesis K-star, and Erdős’s later unpublished positive-density variant. Those are useful literature leads, but the page itself warns that its status reflects the owner’s current belief and may omit literature. No claim here promotes the unpublished argument to a proved result.

Thus a verified lower bound is known, but it does not settle the page’s stronger question about a fixed \(c>0\) and infinitely many \(n\). The maintained page continues to label the problem open. The order of growth in the broad sense is likewise not supplied by the 1936 lower bound alone.

GCL's current state

The repository’s exact registry entry for problem 322 records catalogue status open, triage class FORMAL_SOURCE_REFRESH_REQUIRED, no protected FormalConjectures file, and the next action REFRESH_OR_SOURCE_LOCK_CURRENT_FORMALIZATION_BEFORE_CAMPAIGN_DESIGN. This is a workflow and provenance state; the V2 ledger decides accepted coverage.

The live page now links a FormalConjectures file. The file inspected at upstream commit c892fcbc758f2d978f0f66dbad124bd454b2d619 has blob 5f1f37187cbd1c9f6604284ecd86268fde2d9cbc. It defines representationCount using positive bases bounded by \(n\), asks for each \(k\geq 3\) whether some positive real \(c\) has infinitely many \(n\) with \((n: \mathbb R)^c<\texttt{representationCount}\ k\ n\), and marks the theorem by sorry. This captures the lower-bound question but not the first “what is the order of growth?” question as a separate formal target.

The issue’s admitted FC-GDM-85F86371 snapshot explicitly has no statement file for 322. Therefore the upstream file is current external evidence, not a protected GCL formal source, and its sorry is not a proof or certification. The source packet’s original RESULT/1 citation is authenticated original RESULT/1; no future link is invented here.

What remains

A useful bounded successor task is a source-normalization certificate for \(k=3\). Let \(p_j(n)\) count ordered j-tuples of positive integers whose cubes sum to \(n\), and let \(f_3(n)\) count ordered triples of nonnegative integers as in Erdős’s paper. For every \(n>0\), partition a triple by its number of zero coordinates. The exact identity to prove is

\[ f_3(n)=p_3(n)+3p_2(n)+3\,\mathbf 1_{\{n=x^3\text{ for some }x>0\}}. \]

The input is the scanned 1936 definition and the pinned or explicitly labelled upstream formal definition. The target is a written bijection for the three cases, together with exact integer enumeration for \(1\leq n\leq 1000\) as a regression check. Acceptance requires the symbolic partition proof, the script, and a statement of which convention the resulting lower-bound interface uses. A counterexample to the identity, or an ambiguity about whether representations are ordered, falsifies the package. Stop after this bridge is checked; do not infer the open asymptotic question from the finite table and do not launch a Lean build.

Sources and claim boundaries

The primary mathematical evidence is Erdős (1936), the scanned paper, pp. 133–136. The maintained status and explanatory context are Erdős Problems #322, accessed during this audit; its bibliography also identifies Erdős (1965), Some recent advances and current problems in number theory, and Erdős–Graham (1980), Old and new problems and results in combinatorial number theory. The frozen catalogue evidence is the problems.yaml blob at commit 536408716724a2402bdfa7c1d764bf83d64a8c89, blob 4e60263a80af6207aca494d2e1388e44fa372889. The external formal evidence is the exact 322.lean file at commit c892fcbc758f2d978f0f66dbad124bd454b2d619, blob 5f1f37187cbd1c9f6604284ecd86268fde2d9cbc.

The 1936 lower bound is reported from the primary paper, not reproved here. The finite calculation is exact and reproducible, but finite checks do not establish an infinite statement. The positive/nonnegative base discrepancy, the missing protected formal snapshot, and the sorry theorem are explicit unresolved boundaries. No Lean replay, mathematical certification, publication decision, award claim, or accepted-coverage claim is supplied by this article.

Catalogue and evidence provenance

  • Pinned catalogue source: teorth/erdosproblems at 536408716724.
  • GCL issue: #1592 — protected issue/assignment identity; its labels and results are live, not reproduced automatically here.
  • GCL triage: FORMAL_SOURCE_REFRESH_REQUIRED.
  • Formal-statement coverage: No statement was confirmed in GCL's pinned formal-provider snapshot. Presence does not mean the statement or proof has been checked.
  • External mathematical status: consult current problem page and its dated bibliography.

Back to catalogue · Compendium introduction